Dead time optimization method for hierarchical predictive control based on extended voltage vector

Abstract

The invention discloses a dead time optimization method of hierarchical predictive control based on extended voltage vectors. The method comprises the following steps: firstly, configuring a vector control set of traditional multi-step model predictive current control to be on the basis of six basic voltage vectors, six extended voltage vectors and two zero voltage vectors; then, adopting a hierarchical predictive control strategy to coarsely screen the 14 voltage vectors to obtain a plurality of candidate voltage vectors; and then, constructing a cost function containing a current tracking error term and a dead time penalty term, predicting each candidate voltage vector, generating optimal dead time based on a fine screening result, and adjusting and optimizing a target weight in real time through a dynamic weight factor. The method is high in prediction precision and efficiency, and improves the dynamic performance and adaptive capability of a control system.

Description

Technical Field

[0001] The present invention relates to a dead time optimization method, and particularly to a dead time optimization method based on hierarchical predictive control of extended voltage vectors. Background Art

[0002] Permanent magnet synchronous motors (PMSMs) have been widely used in new energy vehicles, industrial servo systems, aerospace and other fields due to their advantages such as high power density and high efficiency. However, the dead time effect generated during the switching process of inverter power devices has always been a key bottleneck restricting the improvement of system performance. The voltage distortion caused by dead time not only increases current harmonics and electromagnetic torque ripple, but also leads to problems such as increased device losses and excessive electromagnetic interference (EMI). Existing multi-step model predictive current control schemes adopt a 7-vector topology of 6 basic voltage vectors + 1 zero voltage vector. The technical problems are as follows: The 60° angle between adjacent vectors makes it impossible to accurately match the compensation voltage direction near the low load area and the current zero crossing point, resulting in low accuracy of the calculated dead time; and in the scenarios of load mutation or high-speed operation, the fixed vector direction is difficult to track the rapidly changing current slope, leading to an unsatisfactory operating state of the motor. Summary of the Invention

[0003] Object of the Invention: The object of the present invention is to provide a more accurate and efficient dead time optimization method based on hierarchical predictive control of extended voltage vectors.

[0004] Technical Solution: A dead time optimization method based on hierarchical predictive control of extended voltage vectors according to the present invention includes:

[0005] (1) Configuring a vector control set to 6 basic voltage vectors, 6 extended voltage vectors and 2 zero voltage vectors;

[0006] (2) Adopting a hierarchical predictive control strategy, including a rough screening stage and a fine screening stage;

[0007] In the rough screening stage, by constructing a scoring function, several voltage vectors with higher scores are selected from 14 voltage vectors as candidate voltage vectors;

[0008] In the fine screening stage, by constructing a cost function that can comprehensively evaluate the current tracking error and the dead time deviation, and then selecting the candidate voltage vector that minimizes the cost function and its corresponding dead time to generate a corresponding space vector pulse width modulation signal for controlling the operation of the inverter.

[0009] Further, taking any two adjacent basic voltage vectors as boundaries to divide regions, and taking each of the two basic voltage vectors in each region to act for 1 / 2 control period. In this way, 6 extended voltage vectors are obtained, and the amplitude of the extended voltage vector is 0.866 times that of the basic voltage vector.

[0010] Furthermore, the coarse screening stage includes:

[0011] 1) Design a current gradient prediction model

[0012] For the i-th extended voltage vector V i =(V d i , V q i ), calculate the current change in the next cycle:

[0013]

[0014] where Δi d (V i ) is the d-axis current change corresponding to the i-th extended voltage vector; Δi q (V i ) is the q-axis current change corresponding to the i-th extended voltage vector; is the effective d-axis voltage considering dead-time compensation; is the effective q-axis voltage considering dead-time compensation; is the d-axis effective voltage value of the i-th extended voltage vector; is the q-axis effective voltage value of the i-th extended voltage vector; R s is the stator resistance; i d (k) is the d-axis current value at time k; i q (k) is the q-axis current value at time k; w(k) is the rotor electrical angular velocity value at time k; L d is the d-axis inductance; L q is the q-axis inductance; is the permanent magnet flux linkage; T sw is the switching period;

[0015] 2) Construct a scoring function:

[0016]

[0017]

[0018] where score(V i ) is the comprehensive voltage vector score; Δi dq (V i ) is the dq-axis current change under the action of the voltage vector V i ; ||Δi dq || max is the maximum value of the d-q current vector modulus; is the predicted ideal dead time; is the nominal dead time; is the maximum allowable value of the dead time; is the minimum allowable value of the dead time.

[0019] Furthermore,

[0020]

[0021] wherein, is the i-th extended voltage vector of the d-axis; is the i-th extended voltage vector of the q-axis; is the predicted ideal dead time; V dc (k) is the DC bus voltage at time k.

[0022] Furthermore,

[0023]

[0024] wherein, L s is the equivalent inductance of the motor; is the predicted ideal dead time current value; K safe is the safety factor.

[0025] Furthermore,

[0026] Furthermore, all voltage vectors are sorted in descending order of Score to generate a candidate sequence V sorted = Sort{V1, V1,... V 14}; when it is detected that the angular acceleration of the motor is not higher than 1000 rad / s 2 , the 7 voltage vectors with the highest scores are selected as candidate voltage vectors.

[0027] Furthermore, when it is detected that the angular acceleration of the motor is higher than 1000 rad / s 2 , the 10 voltage vectors with the highest scores are selected as candidate voltage vectors.

[0028] Furthermore, the fine screening stage includes

[0029] 1) Predict the current for the next three switching cycles for each candidate voltage vector V i :

[0030]

[0031]

[0032] wherein, the effective voltage is:

[0033]

[0034] 2) Construct the cost function J to comprehensively evaluate the current tracking error and dead-time deviation:

[0035]

[0036] Among them, is the reference current vector in the d-q coordinate system; is the reference current vector at the nth step; is the predicted current vector at the nth step; λ is the dynamic weight factor; is the predicted dead time at the nth step;

[0037] Select the voltage vector with the minimum cost function and its corresponding dead time to generate the corresponding space vector pulse width modulation signal:

[0038]

[0039] Among them, V opt is the optimal voltage vector; is the optimal dead time; J(V i ) is the cost function value.

[0040] Furthermore,

[0041]

[0042] Among them, λ0 is the reference weight; THD real is the total harmonic distortion rate of the real-time current; THD threshold = 0.05; is the real-time angular velocity of the motor.

[0043] Through the design of the dynamic weight factor, the present invention improves the dynamic parameter adjustment and adaptive ability, and enhances the dynamic performance and adaptive ability of the control system.

[0044] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: The present invention adopts an extended voltage vector and hierarchical prediction collaborative control architecture. Based on the traditional 6+1 vector, 7 sub-division vectors are newly added through spatial subdivision to form a dense voltage vector coverage network. Combined with the hierarchical prediction algorithm, the defect of "step-by-step" compensation in traditional dead-time optimization is solved, making the dead-time optimization more accurate. The hierarchical prediction algorithm performs hierarchical optimization based on the extended voltage vector, greatly reducing the computational amount of dead-time optimization and ensuring the computational efficiency. Description of the Drawings

[0045] Figure 1 is the extended voltage vector diagram in the embodiment of the present application;

[0046] Figure 2It is the rotational speed diagram of the hierarchical predictive control strategy from no-load start to 1600 r / min before and after improvement in the embodiments of the present application;

[0047] Figure 3 It is the simulation waveform diagram of the d-axis current when the hierarchical predictive control strategy in the embodiments of the present application is no-load and the rotational speed is 1600 r / min before and after improvement. Specific embodiments

[0048] The present invention will be further described below with reference to the accompanying drawings.

[0049] The embodiments of the present invention provide a dead-time optimization method for hierarchical predictive control based on an extended voltage vector, including the following steps:

[0050] (1) As Figure 1 shown, configure the vector control set of the traditional multi-step model predictive current control into 6 basic voltage vectors, 6 extended voltage vectors, and 2 zero voltage vectors;

[0051] Take any two adjacent basic voltage vectors as boundaries to divide the region, and let each of the two basic voltage vectors in each region act for 1 / 2 control period (T s / 2). In this way, 6 extended voltage vectors are obtained, and the amplitude of the extended voltage vector is 0.866 times that of the basic voltage vector.

[0052] (2) Adopt a hierarchical predictive control strategy, including a rough screening stage and a fine screening stage;

[0053] In the rough screening stage, construct a scoring function, and then select several voltage vectors with higher scores from 14 voltage vectors as candidate voltage vectors. Specifically, it includes:

[0054] 1) Design a current gradient prediction model

[0055] For the i-th extended voltage vector V i =(V d i , V q i ), calculate the current change in the next period:

[0056]

[0057] Among them, Δi d (V i ) is the d-axis current change amount corresponding to the i-th extended voltage vector; Δi q (V i ) is the q-axis current change amount corresponding to the i-th extended voltage vector; is the effective d-axis voltage considering dead-time compensation; is the effective q-axis voltage considering dead-time compensation; is the d-axis effective voltage value of the i-th extended voltage vector; is the q-axis effective voltage value of the i-th extended voltage vector; R s is the stator resistance; i d (k) is the d-axis current value at time k; i q (k) is the q-axis current value at time k; w(k) is the rotor electrical angular velocity value at time k; L d is the d-axis inductance; L q is the q-axis inductance; is the permanent magnet flux linkage; T sw is the switching period.

[0058]

[0059] Among them, is the i-th extended voltage vector of the d-axis; is the i-th extended voltage vector of the q-axis; is the predicted ideal dead time; V dc (k) is the DC bus voltage at time k.

[0060] Calculate the ideal dead time corresponding to each vector:

[0061]

[0062] Among them, L s is the equivalent inductance of the motor; is the predicted ideal dead time current value; K safe is the safety factor.

[0063] 2) Construct a scoring function:

[0064]

[0065]

[0066] Among them, score(V i ) is the comprehensive voltage vector score; Δi dq (V i ) is the change in dq-axis current under the action of voltage vector V i ; ||Δi dq || max is the maximum value of the d-q current vector modulus; is the predicted ideal dead time; is the nominal dead time, taking 100 ns; is the maximum allowable value of the dead time; is the minimum allowable value of the dead time.

[0067] Arrange all voltage vectors in descending order of Score to generate a candidate sequence V sorted = Sort{V1, V1,... V 14}; When the angular acceleration dw / dt of the motor is detected to be no higher than 1000 rad / s 2 , select the top 7 voltage vectors with higher scores as candidate voltage vectors. When the angular acceleration dw / dt of the motor is detected to be higher than 1000 rad / s 2 , select the top 10 voltage vectors with higher scores as candidate voltage vectors. This situation is an abnormal working condition.

[0068] In the fine screening stage, a cost function that can comprehensively evaluate the current tracking error and dead time deviation is constructed, and then the candidate voltage vector that minimizes the cost function and its corresponding dead time are selected to generate the corresponding space vector pulse width modulation signal for controlling the operation of the inverter. Specifically, it includes:

[0069] 1) Predict the current for the next three switching cycles for each candidate voltage vector V i :

[0070]

[0071] Among them, the effective voltage is:

[0072]

[0073] 2) Construct a cost function J to comprehensively evaluate the current tracking error and dead time deviation:

[0074]

[0075] Among them, is the reference current vector in the d-q coordinate system; is the reference current vector at the nth step; i dq (k + n) is the predicted current vector at the nth step; λ is the dynamic weight factor; is the predicted dead time at the nth step.

[0076]

[0077] Among them, λ0 is the reference weight, taking 0.5; THD real is the total harmonic distortion rate of the real-time current; THD threshold = 0.05; is the real-time angular velocity of the motor; λ ∈ [0.5, 3.0].

[0078] Select the voltage vector with the minimum cost function and its corresponding dead time to generate the corresponding space vector pulse width modulation signal:

[0079]

[0080] Among them, V opt is the optimal voltage vector; is the optimal dead time; J(V i ) is the cost function value.

[0081] Figure 2 is the speed diagram of the hierarchical predictive control strategy before and after improvement in the embodiment of the present application during no-load starting to 1600 r / min. Figure (a) is the traditional hierarchical predictive control strategy, and Figure (b) is the hierarchical predictive control strategy based on the extended voltage vector. Starting from no-load to the given speed of 800 rad / min, the given speed suddenly changes to 1600 rad / min at 0.08 s, and the given speed suddenly changes to 1300 rad / min at 0.16 s. It can be seen from the simulation waveforms that the hierarchical predictive control strategy based on the extended voltage vector can quickly follow the speed command, and the two methods have similar dynamic performances during the starting, rising and falling, and decelerating processes.

[0082] Figure 3 is the simulation waveform diagram of the d-axis current of the hierarchical predictive control strategy before and after improvement in the embodiment of the present application when it is no-load and the speed is 1600 r / min. Figure (a) is the traditional hierarchical predictive control strategy, and Figure (b) is the hierarchical predictive control strategy based on the extended voltage vector. The two strategies are both the simulation waveform diagrams of the d-axis current obtained under no-load and the given speed of 1600 rad / min. It can be seen that the d-axis pulsation obtained by the hierarchical predictive control strategy based on the extended voltage vector is significantly smaller than that of the traditional hierarchical predictive control strategy.

Claims

1. A dead-time optimization method based on hierarchical predictive control of extended voltage vectors, characterized in that Including: (1) Configure the vector control set as 6 basic voltage vectors, 6 extended voltage vectors, and 2 zero voltage vectors; (2) Adopt a hierarchical predictive control strategy, including a rough screening stage and a fine screening stage; In the rough screening stage, by constructing a scoring function, and then select several voltage vectors with higher scores from 14 voltage vectors as candidate voltage vectors; In the fine screening stage, by constructing a cost function that can comprehensively evaluate the current tracking error and the dead-time deviation, and then select the candidate voltage vector that minimizes the cost function and its corresponding dead-time to generate the corresponding space vector pulse width modulation signal for controlling the operation of the inverter.

2. The dead-time optimization method based on hierarchical predictive control of extended voltage vectors according to claim 1, wherein Divide the area with any two adjacent basic voltage vectors as the boundary, and let each of the two basic voltage vectors in each area act for 1 / 2 control cycle. In this way, 6 extended voltage vectors are obtained, and the amplitude of the extended voltage vector is 0.866 times that of the basic voltage vector.

3. The dead - time optimization method based on hierarchical predictive control of extended voltage vectors according to claim 1, wherein The rough screening stage includes: 1) Design a current gradient prediction model For the i-th extended voltage vector V i =(V d i , V q i ), calculate the current change in the next cycle: where, Δi d (V i ) is the change in d-axis current corresponding to the i-th extended voltage vector; Δi q (V i ) is the change in q-axis current corresponding to the i-th extended voltage vector; is the effective d-axis voltage considering dead-time compensation; is the effective q-axis voltage considering dead-time compensation; is the d-axis effective voltage value of the i-th extended voltage vector; is the q-axis effective voltage value of the i-th extended voltage vector; R s is the stator resistance; i d (k) is the d-axis current value at time k; i q (k) is the q-axis current value at time k; w(k) is the rotor electrical angular velocity value at time k; L d is the d-axis inductance; L q is the q-axis inductance; is the permanent magnet flux linkage; T sw is the switching period; 2) Construct a scoring function: Among them, score(V i ) is the comprehensive score of the voltage vector; Δi dq (V i ) is the change in the dq-axis current under the action of the voltage vector V i ; ||Δi dq || max is the maximum value of the d-q current vector modulus; is the predicted ideal dead time; is the nominal dead time; is the maximum allowable value of the dead time; is the minimum allowable value of the dead time.

4. The dead-time optimization method based on hierarchical predictive control with extended voltage vectors according to claim 3, characterized in that Among them, is the i-th extended voltage vector on the d-axis; is the i-th extended voltage vector on the q-axis; is the predicted ideal dead time; V dc (k) is the DC bus voltage at the k-th moment.

5. The dead-time optimization method based on hierarchical predictive control with extended voltage vectors according to claim 4, characterized in that Among them, L s is the equivalent inductance of the motor; is the predicted ideal dead - zone current value; K safe is the safety factor.

6. The dead-time optimization method based on hierarchical predictive control of extended voltage vectors according to claim 5, characterized in that 7. The dead-time optimization method based on hierarchical predictive control with extended voltage vectors according to claim 6, characterized in that Arrange all voltage vectors in descending order of Score to generate a candidate sequence V sorted = Sort{V1, V1,... V 14}; When it is detected that the angular acceleration of the motor is not higher than 1000 rad / s 2 Select the top 7 voltage vectors with higher scores as candidate voltage vectors.

8. The dead-time optimization method based on hierarchical predictive control of extended voltage vectors according to claim 7, characterized in that When the angular acceleration of the motor is detected to be higher than 1000 rad / s 2 select the top 10 voltage vectors with the highest scores as candidate voltage vectors.

9. The dead-time optimization method for hierarchical predictive control based on extended voltage vectors according to claim 6, characterized in that The fine screening stage includes 1) For each candidate voltage vector V i perform current prediction for the next three switching cycles: Wherein, the effective voltage is: 2) Construct a cost function J to comprehensively evaluate the current tracking error and the dead-time deviation: Among them, is the reference current vector in the d-q coordinate system; is the reference current vector at the nth step; i dq (k + n) is the predicted current vector at the nth step; λ is the dynamic weight factor; is the predicted dead time at the nth step; Select the voltage vector with the minimum cost function and its corresponding dead-time to generate the corresponding space vector pulse width modulation signal: Among them, V opt is the optimal voltage vector; is the optimal dead time; J(V i ) is the cost function value.

10. The dead-time optimization method based on hierarchical predictive control with extended voltage vectors according to claim 9, characterized in that Among them, λ0 is the reference weight; THD real is the total harmonic distortion rate of the real-time current; THD threshold = 0.05; is the real-time angular velocity of the motor.

Images